Dee et al. BMC Veterinary Research (2016) 12:51 DOI 10.1186/s12917-016-0674-z

RESEARCH ARTICLE Open Access

Modeling the transboundary risk of feedingredients contaminated with porcineepidemic diarrhea virus

Scott Dee1*, Casey Neill1, Aaron Singrey2, Travis Clement2, Roger Cochrane3, Cassandra Jones3, Gilbert Patterson4,Gordon Spronk1, Jane Christopher-Hennings2 and Eric Nelson2

Abstract

Background: This study describes a model developed to evaluate the transboundary risk of PEDV-contaminatedswine feed ingredients and the effect of two mitigation strategies during a simulated transport event from China tothe US.

Results: Ingredients imported to the USA from China, including organic & conventional soybeans and meal, lysinehydrochloride, D-L methionine, tryptophan, Vitamins A, D & E, choline, carriers (rice hulls, corn cobs) and feed gradetetracycline, were inoculated with PEDV. Control ingredients, and treatments (ingredients plus a liquid antimicrobial(SalCURB, Kemin Industries (LA) or a 2 % custom medium chain fatty acid blend (MCFA)) were tested. The modelran for 37 days, simulating transport of cargo from Beijing, China to Des Moines, IA, US from December 23, 2012 toJanuary 28, 2013. To mimic conditions on land and sea, historical temperature and percent relative humidity (% RH)data were programmed into an environmental chamber which stored all containers. To evaluate PEDV viability overtime, ingredients were organized into 1 of 4 batches of samples, each batch representing a specific segment oftransport. Batch 1 (segment 1) simulated transport of contaminated ingredients from manufacturing plants inBeijing (day 1 post-contamination (PC)). Batch 2 (segments 1 and 2) simulated manufacturing and delivery toShanghai, including time in Anquing terminal awaiting shipment (days 1–8 PC). Batch 3 (segments 1, 2 and 3)represented time in China, the crossing of the Pacific and entry to the US at the San Francisco, CA terminal(day 1–27 PC). Batch 4 (segments 1–4) represented the previous events, including transport to Des Moines, IA (days1–37 PC). Across control (non-treated) ingredients, viable PEDV was detected in soybean meal (organic andconventional), Vitamin D, lysine hydrochloride and choline chloride. In contrast, viable PEDV was not detected inany samples treated with LA or MCFA.

Conclusions: These results demonstrate the ability of PEDV to survive in a subset of feed ingredients using amodel simulating shipment from China to the US. This is proof of concept suggesting that contaminated feedingredients could serve as transboundary risk factors for PEDV, along with the identification of effective mitigationoptions.

Keywords: Transboundary, Porcine, Epidemic, Diarrhea, Virus, Antimicrobial, Ingredient, Lysine, Soybean meal,Choline

* Correspondence: sdee@pipevet.com1Pipestone Applied Research, Pipestone Veterinary Services, 1300 Box188Hwy 75 S, Pipestone, MN 56164, USAFull list of author information is available at the end of the article

© 2016 Dee et al. Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Table 1 Summary of quantities (kg) of representative feedingredients imported to the US from China in 2013 and 2014a

Ingredient 2013 (kg) 2014 (kg)

Soybeans (conventional and organic) 89,469,255 70,940,336

Lysine hydrochloride 21,469,477 5,382,454

Soybean meal (conventional and organic) 6,632,960 7,214,863

Feed-grade tetracycline 1,411,251 1,418,012

D-L methionine 377,495 1,790,280

Vitamin E 12,640 14,194

Choline chloride 11,284 23,572

Vitamin A 1156 1094

Vitamin D 1043 1052a: Source = US Government Harmonized Tariff Schedule

Dee et al. BMC Veterinary Research (2016) 12:51 Page 2 of 12

BackgroundPorcine epidemic diarrhea virus (PEDV) is an envelopedsingle-stranded positive sense RNA virus belonging tothe Order Nidovirales, the family Coronaviridae and thegenus Alphacoronavirus [1]. Following detection in theUS swine population during May, 2013 the virus spreadrapidly throughout the country [2]. During the initialoutbreak, the American Association of Swine Veterinar-ians, the National Pork Producers Council and theUSDA Center for Epidemiology and Animal Health con-ducted an epidemiological investigation involving caseand control herds [3]. Of the 100 variables surveyed, 7were significantly associated (p < 0.05) with havinghigher odds of acquiring PED, with all associated withthe process of feeding animals [3]. In 2014, the risk ofcontaminated feed and feed ingredients was confirmed,when studies describing the ability of PEDV to survivefor extended periods in feed (7 days in dry feed and28 days in wet feed when stored at room temperature),and proof of concept that contaminated complete feedand feed ingredients could serve as vehicles for trans-mission to naïve pigs [4–6]. Shortly thereafter, the trans-mission of PEDV via ingestion of contaminatedcomplete feed was validated, along with calculation ofthe minimum infectious dose of the virus in feed [7].This information heightened the awareness of the needfor strategies to mitigate the risk of contaminated feed,through thermal processing or chemical treatment. Inregards to the latter approach, data are now availabledemonstrating the ability of a liquid antimicrobial prod-uct containing formaldehyde and propionic acid or amedium chain fatty acid blend to successfully degradePEDV RNA in contaminated feed and prevent infection[8, 9].Despite these advances at the domestic level, the risk

of PEDV infection at the global level still exists, as thesource of the initial PEDV introduction to the US re-mains unidentified. During the widespread epidemic, therole of feed in transboundary spread of the virus wasdownplayed, based on lack of data supporting survival ofPEDV in feed ingredients over time and under condi-tions representative of trans-oceanic transport. Recently,extended survival of PEDV in individual feed ingredientsunder wintertime ambient conditions has been reported[10]. Most notably, survival of PEDV was demonstratedin soybean meal for 180 days, along with evidence ofvirus survival in lysine hydrochloride, choline chloride,DDGS and several porcine by-products for at least30 days [10]. One feature that was consistent across sev-eral of these ingredients was that they are imported tothe US from China. As the original PEDV detected inthe US is closely related to a Chinese variant [11], itraises the question whether contaminated feed ingredi-ents imported from China could have served as a source

for viral entry to the US in 2013. Table 1 is a summaryof data derived from the US International Trade Com-mission Harmonized Tariff Schedule website (www.hs.u-sitc.gov) which is a publicly available database thatprovides a transaction of specific trade commodities be-tween the US and its international trading partners.These data provide insight into the types of and quan-tities of ingredients that entered the US from China dur-ing the height of the PED epidemic in 2013 and 2014(G. Patterson, personal communication, May 2015). Thiscollective information justifies renewed investigationinto the transboundary risk of PEDV and the potentialrole that contaminated feed ingredients could play in thespread of disease during the process of trans-oceanicshipping. Therefore, the purpose of this study was to de-velop a model to evaluate the transboundary risk ofPEDV-contaminated swine feed ingredients during asimulated shipment from China to the US, as well as testthe effect of two mitigation strategies. The study wasbased on the hypothesis that while select non-treated in-gredients could provide a protective effect on PEDV sur-vival, mitigation would reduce risk.

MethodsDesign of shipping modelRoute and timetableBased on the predominance of facilities processing agri-cultural feed ingredients for export in the eastern regionof China, i.e., Shandong, Jilin, Henan, Hebei and Liao-ning provinces (www.alibaba.com), and that the PEDVstrain initially detected in the US most closely resembleda variant from the province of Anhui [11], the city ofBeijing was designated as the starting point for themodel. Here it was assumed that PEDV contaminationof select ingredients would occur, either at the manufac-turing plant or post-processing [4]. In an effort to deter-mine if PEDV could be delivered in a viable state fromBeijing to a major pork production region in the US,

Dee et al. BMC Veterinary Research (2016) 12:51 Page 3 of 12

Des Moines, IA was selected as the final destination.Based on these assumptions, a commercial website(SeaRates.com) was used to develop a representativeroute and timetable. Specifically, it was modeled thatcontaminated ingredients would travel for 1 day fromBeijing to the Anquing terminal in Shanghai, where theywould be held for 7 days in preparation for shipment tothe US. Cargo would then travel across the PacificOcean over a 17 day period and enter the US at the portof San Francisco. Following a 7 day period to clear cus-toms, it would then be transported for 2 days via Inter-state 80 to Des Moines, where it would remain for3 days, for a total transit period of 37 days (Table 2).

Compilation of environmental dataOnce the 37-day timetable had been established, it wasdecided to model the shipping event over the period ofDecember 23, 2012 to January 28, 2013. This decisionwas based on the timing of the initial detection of PEDVin the US (April 2013) and the availability of data sum-marizing temperature and % RH in shipping containerstraveling from Asia to the US during the period of De-cember 31, 2012 to January 16, 2013 [12]. Using this in-formation, we designed a temperature and % RH curvefor the oceanic segment of the study. In addition, weaccessed historical meteorological data for the land seg-ments of the model using Weather Underground(www.wunderground.com) encompassing the periods ofDecember 23–30, 2012 and January 17–27, 2013, whichwere paired in conjunction with the oceanic transportperiod. To simulate the effect of daily fluctuation we col-lected historical temperature data at 4 designated timeseach day (6 AM, 12 PM, 6 PM, 12 AM) and at 3 desig-nated times each day for % RH (8 AM, 12 PM, 4 PM).All data were then entered into the computer of an en-vironmental chamber used to house samples during themodel shipping period.

Processing of feed ingredientsA panel of 14 swine feed ingredients known to beimported to the US from China were selected for thisstudy, including organic & conventional soybeans andsoybean meal, lysine hydrochloride, D-L methionine,tryptophan, Vitamins A, D & E, choline chloride, two

Table 2 Description of locations and events involved in the 37-day

Day Location

1 Beijing

2–8 Anquing terminal, Shanghai

9–25 Pacific ocean

26–32 San Francisco terminal, CA, US

33–34 Interstate 80

35–37 Des Moines, IA

ingredient carriers (rice hulls or corn cobs) and feedgrade tetracycline. In regards to the soy-based products,the guaranteed analysis of conventional meal indicated a48 % crude protein, 1 % fat and 3 % fiber while the or-ganic product had lower protein (44 %) and higher fatand fiber (7.5 and 6.5 %, respectively). Furthermore, theprocess of manufacturing organic soybean meal was voidof chemical (hexane) use and no chemical fertilizer hadbeen used during the soybean growing period. Ingredi-ents were screened by PCR to insure a PEDV-negativestatus prior to the onset of the study. The treatments se-lected for the study included a liquid antimicrobial (LA)(SalCURB®, Kemin Industries, Des Moines, IA USA) or amedium chain fatty acid blend (MCFA). SalCURB® is apremix of aqueous formaldehyde solution 37 % (formaintenance of complete animal feeds or feed ingredi-ents Salmonella-negative for up to 21 days) and propio-nic acid (as a chemical preservative for control of moldin feed or feed ingredients). While SalCURB® provideseffective Salmonella control for up to 21 days, it is notapproved for use by the U.S. Food & Drug Administra-tion or the U.S. Department of Agriculture as a treat-ment for PEDV. The second treatment, MCFA, was a2 % custom medium chain fatty acid blend of caproic,caprylic and capric acids, blended at a 1:1 ratio [9]. Con-trol ingredients were treated with sterile saline.

Sample managementSamples were organized into 1 of 4 identical batches,each representing a specific segment of the 37-day ship-ping period. Batch 1 (segment 1) was designed to repre-sent the transport of contaminated ingredients frommanufacturing plants in Beijing to the ShanghaiAnquing terminal (day 1 post-contamination (DPC)).Batch 2, a compilation of segments 1 and 2, simulatedmanufacturing and delivery to Shanghai, as well as timein the Anquing terminal awaiting shipment (1–8 DPC).Batch 3, a compilation of segments 1, 2 and 3 repre-sented time in China, the crossing of the Pacific and ar-rival to the US at the San Francisco, CA terminal (1–27DPC). Finally, batch 4 was a compilation of segments 1–4, thereby representing the entire process, includingtransport to and storage in, Des Moines, IA (1–37 DPC).On designated days post-contamination, a batch of

shipping model period

Events

Ingredient manufacturing & contamination

Transport to Shanghai, awaiting shipment to US

Departure from China, crossing the ocean

Entry to the US, clearing customs

Transport from California to Iowa

Storage

Dee et al. BMC Veterinary Research (2016) 12:51 Page 4 of 12

samples was removed from the environmental chamberand submitted for testing. Specifically, batch 1 was sub-mitted on 1 DPC, batch 2 on 8 DPC, batch 3 on 27 DPCand batch 4 on 37 DPC. In other words, the same sam-ple was not repeatedly opened and tested, but rather anew batch of samples was submitted on a designatedday. This method of sample management would insurethat all sample containers remained sealed from the timethey were inoculated until the time they were tested atthe lab, minimizing the risk of cross-contamination, aswell as enhancing repeatability of results as several seg-ments were replicated across batches. Finally, to increasestatistical power, each of the 4 batches contained 2 repli-cates of each ingredient in the control group and 2 repli-cates of each ingredient within each treatment group,for a total of 90 samples per batch.

Sample contamination proceduresA goal of the study was to limit the variability to thelevel of the ingredient; therefore, the same quantity ofingredient, the same container type and the same envir-onmental settings were used and samples were contami-nated equally. To initiate this process, 30 g of eachingredient were added to food storage containers (OxoTot Baby Blocks, Oxo International, El Paso, TX, USA)to simulate a shipping container [10]. Ingredients in thenon-treated control group were treated with 0.1 mL ofsterile saline. Ingredients in the LA group were treatedwith 0.1 mL of product, based on an inclusion rate of3 kg/ton of complete feed. Ingredients in the MCFAgroup were treated with 0.6 g of product based on a 2 %inclusion rate. Individual treatments and saline placebowere added to the designated samples using separate tu-berculin syringes. To promote proper mixing, the feedwas stirred manually for 10 clockwise rotations and 10counter-clockwise rotations using individual wooden ap-plicator sticks per ingredient. Following mixing, each in-dividual container was manually shaken vigorously (50times in a 10 s period). All samples were then inoculatedwith 2 mL PEDV (passage 18, Ct = 17.15, total dose491,520 FFU) and mixed as described. This quantity ofPEDV was selected in an effort to provide a final meanCt value in feed ingredient of approximately 25 (range =19–30) following mixing, based on data from actual fieldcases of PEDV-contaminated feed, a challenge level usedin published studies [5, 8, 10].

ControlsFor the purpose of negative controls, 30 g samples ofPEDV-negative complete feed were inoculated with ster-ile saline. Duplicate negative controls were included ineach of the 4 batches, across the control and treatmentgroups. For the purpose of positive controls, duplicate5 mL samples of stock PEDV in MEM (minimum

essential media, Gibco, ThermoFisher Scientific, Wal-tham, MA, USA) were included in containers withineach batch of ingredients in both control and treatmentgroups.

Sample storageOnce prepared, samples were stored in the environmen-tal chamber. This instrument (Model 9005 L, SheldonManufacturing Inc., Cornelius, OR) had a programmabletemperature range of 4–210 C and a RH range of 40–95 %. In order for ingredients to be exposed to ambientair within the chamber, two holes, 0.318 cm in diameterwere drilled into each plastic container (Fig. 1). Usingthe data from the historical temperature and % RHcurve described above, the chamber computer was pro-grammed to simulate fluctuation over time as previouslydescribed.

Diagnostic proceduresAll diagnostic testing was conducted using protocols de-veloped and validated by the South Dakota State Univer-sity (SDSU) Animal Disease Research and DiagnosticLaboratory (ADRDL). Samples were submitted by codeto the laboratory, so personnel were blinded as to batch,treatment versus non-treatment and ingredient type. Itwas originally planned to test all samples by PCR andvirus isolation (VI), followed by the use of swine bio-assay for samples determined to be PCR positive but VInegative.

Extraction of RNAThe MagMAXTM 96 Viral Isolation Kit (Life Technolo-gies, Waltham MA, USA) was used to obtain viral RNAfrom the samples, as described in the instructions pro-vided (1836 M Revision F). A 175-μl volume of samplewas used for the extraction. The magnetic bead extrac-tions were completed on a Kingfisher96 instrument(Thermo Scientific, Waltham MA, USA).

Real-time PCRA commercially available real-time, single tube RT-PCRmultiplex assay for the detection of PEDV, porcine delta-coronavirus (PDCoV) and transmissible gastroenteritisvirus (TGEV) was used in this study per kit instruction(Tetracore, Rockville, MD, USA). Briefly, 7 μl of the ex-tracted RNA was added to 18 μl of the master mix. Theone-step real-time RT-PCR amplification conditionsstarted with 15 min at 48 °C, followed by 2 min at 95 °C.The final cycles consisted of 5 s at 95 °C and then 40 sat 60 °C (data collection step). The program was run for38 cycles (Cycle time) with PEDV positive results indi-cated at ≤ 38 cycles. Positive and negative controls wereincluded on each run. All amplification was completedon the ABI7500 instrumentation (Austin, TX, USA).

Fig. 1 Environmental chamber and sample containers. This figure depicts the environmental chamber loaded with sample containers as well as aclose-up view of a container. Note the two 0.318 cm holes (red circle) drilled in the container to allow for contact of the ingredient with thechamber environment

Dee et al. BMC Veterinary Research (2016) 12:51 Page 5 of 12

PEDV stock virus propagationFor PEDV propagation, Vero 76 cells (ATCC CRL-1587)were maintained in MEM plus 10 % fetal bovine serumand antibiotics. Three-day old confluent monolayers ofVero 76 cells in 150 cm2 flasks were washed 3 times withserum free minimum essential media (MEM) prior to in-oculation. Monolayers were infected at ~0.1 moi of PEDVin MEM containing 2 ug/ml TPCK treated trypsin, incu-bated at 37 °C for approximately 48 h until obvious CPEwas apparent. Flasks were frozen at −80 °C until needed.

Virus isolationOnce feed ingredient samples were tested for PEDV viaPCR, the residual samples were tested for presence of vi-able virus. Samples were diluted in MEM containing2 μg/ml TPCK-treated trypsin with a starting dilution of1:2 and were two-fold serially diluted. Diluted sampleswere then added to washed confluent monolayers ofVero-76 cells in 96-well plates and incubated for 1 h at37 °C. Plates were again washed and trypsin media re-placed. After 24 h at 37 °C, plates were fixed with 80 %acetone and stained with FITC conjugated mAb SD6-29to allow visualization of infected cells. Virus concentra-tion was determined by calculating FFU/ml based on thenumber of fluorescent foci present in wells at selecteddilutions using a previously published method adaptedto PEDV [13]. Personnel reading the plates were blindedto the type of sample and the time of sampling.

Swine bioassayFacilities and source of animalsThe purpose of the swine bioassay was to determinewhether viable PEDV was present in any feed ingredient

sample that had tested positive on PCR but negative onVI. This study was conducted in a Biosafety Level 2+room at the Animal Resource Wing (ARW) at South Da-kota State University. All procedures involving animalsthroughout the study were performed under the guid-ance and approval of the SDSU Institutional AnimalCare and Use Committee. Animals (n = 24, 5 day oldpiglets) were sourced from a PEDV-naïve herd and weretested on arrival to the ARW via blood sampling andcollection of rectal swabs from each pig. Prior to animalarrival, all rooms (walls, ceilings, floors and drains) weremonitored for the presence of PEDV by PCR using sam-pling procedures previously described [5, 8]. Piglets werehoused in one of 6 stainless steel gnotobiotic unitsmeasuring 0.6 m W × 1.2 m L × 0.6 m H. Units were di-vided into 4 semi-isolated housing units, allowing for 4piglets per unit with individual feeding arrangements.Flooring consisted of an open weave rubberized mat ona perforated stainless steel grate raised 10 cm for wastecollection. Each unit was covered with an inflatable 20mil plastic canopy and fitted with 2 pair of dry-boxgloves for feeding and procedures inside the canopy.Each canopy was secured and sealed with duct tape andratchet straps to the unit. Ventilation was supplied by anelectric fan maintaining sufficient positive pressure in-side the canopy to keep the canopy inflated above theunit. Incoming and outgoing air to each unit was HEPA-filtered. Each unit was initially sterilized using 47 %aerosolized formalin, and allowed to dissipate for 2 weeksprior to introduction of the animals. All incoming andoutgoing materials needed during the study (eg. swabs,injectable medication, bleeding supplies) were passedthrough an air-tight stainless steel port and sterilized

Dee et al. BMC Veterinary Research (2016) 12:51 Page 6 of 12

using 5 % peracetic acid before entering or exiting theport.

Preparation of bioassay inoculumThe stainless steel unit served as the experimental unit;therefore, all 4 piglets in each unit received the same in-gredient. To assess PEDV survivability throughout the en-tire model period, samples from batch 4 were tested,including both treated and non-treated ingredients. Re-garding non-treated samples, all PCR-positive and VI-negative samples in batch 4 were tested. In regards to LA-treated and MCFA-treated ingredients, treated equivalentsof non-treated samples that were VI or swine bioassaypositive, again from batch 4 samples were tested. Forpreparation of the inoculums, 60 g of each specific ingre-dient was mixed with 50 mL of sterile PBS in a 250 mLcentrifuge tube, inverted 10 times to mix and vortexed for2 min. The suspension was then centrifuged at 5200 g for15 min, supernatant decanted and tested by PCR prior topiglet inoculation. Each pig in the unit received 1 mL ofthe designated inoculum orally via syringe and observedfor a 7 day period. To minimize the number of animalsneeded for the study, pigs that were confirmed negativeafter 7 DPI would be inoculated with a different ingredi-ent. A negative control unit was included in the design,with these pigs receiving sterile saline PO.

Table 3 Summary of PCR Ct data mean & (SD) across treatedand non-treated groups on day 1 and 37 of the study

Group Day 1 mean (SD) Day 37 mean (SD)

Piglet testingFollowing inoculation, the PEDV status of each group ofpiglets was monitored [5, 8]. On a daily basis, ARWpersonnel inspected animals for clinical signs of PEDand collected rectal swabs (Dacron swabs, Fisher Scien-tific, Franklin Lakes, NJ, USA) from each pig, startingwith the negative control unit. Showers were taken uponentry to the rooms and room-specific coveralls, footwear,hairnets, gloves and P95 masks (3 M, St. Paul, MN USA)were worn. In addition, each room was ventilated indi-vidually and HEPA filtration for both incoming and out-going air was employed per room. If clinically affectedanimals were observed, swabs of diarrhea and/or vomit-ing, in conjunction with the daily rectal swab were col-lected. Swabs were submitted to the SDSU ADRDL andtested by PCR. If PEDV was diagnosed in a specific unit,all animals were swabbed, humanely euthanized withintravenous sodium pentobarbital, the small intestinaltracts submitted for PCR testing, units were cleaned andsanitized as described and re-stocked with new piglets asneeded.

Control ingredients 22.9a (2.4) 23.1a (3.5)

LA-treated ingredients 24.5a (2.4) 32.5b (3.9)

MCFA-treated ingredients 24.2a (4.2) 25.5a (3.7)

Values with different superscripts (a/b) are significantly different at p < 0.05

Data analysisDescriptive statistics, T-test and ANOVA were used toanalyze data, when applicable.

ResultsSample sizeA total of 360 feed ingredient samples were used for thisstudy.

PCRAll feed ingredient samples were PCR negative on day 0of the study. Successful PEDV inoculation was con-firmed, as all day 1 samples were PCR-positive. Resultsof PCR testing of treated and non-treated ingredients on1 and 37 DPC are summarized in Table 3. The mean Ctof LA-treated samples was 24.5 (SD = 2.4) on 1 DPC and32.5 (SD = 3.9) on 37 DPC (p < 0.0001). The mean Ct ofMCFA-treated samples was 24.2 (SD = 4.2) on 1 DPCand 25.5 (SD = 3.7) on 37 DPC (p = 0.25). The mean Ctvalues across non-treated ingredients on day 1 was 22.9(SD = 2.4) and 23.1 (SD = 3.5) on day 37 (p = 0.34). At 1DPC, the mean Ct values of all 3 groups (non-treated,LA-treated and MCFA-treated) were not significantlydifferent (p = 0.14); however, at 37 DPC, the mean Ct ofLA-treated samples was significantly higher (p < 0.0001)than the mean Ct of MCFA-treated and non-treatedsamples.

Virus isolationA summary of VI data is provided in Table 4. ViablePEDV was recovered from organic and conventionalSBM, lysine and Vitamin D across all 4 batches of non-treated samples, including batch 3 representing entry tothe US at the San Francisco terminal and batch 4, repre-senting shipment to and storage in Des Moines. Noother samples harbored viable virus beyond batch 2(Beijing and Shanghai segments) including the PEDVstock virus control. Multiple samples were VI negativeon 1 DPC. All negative control samples and all LA-treated or MCFA-treated ingredients were VI negativeacross all batches.

Swine bioassaySamples selected for swine bioassay testing consistedof treated and non-treated ingredients from batch 4.The non-treated ingredients tested included thosewhich were PCR- positive and VI-negative, specificallyVitamins A & E, tryptophan, D-L methionine, soy-beans (organic and conventional), and choline

Table 4 Summary of mean PCR Ct and FFU/mL across all 4 batches of non-treated ingredients

Ingredient Mean Mean Mean Mean

Ct/FFU titer Ct/FFU titer Ct/FFU titer Ct/FFU titer

Batch 1 Batch 2 Batch 3 Batch 4

Soybeans (organic) 20.76/5120 23.27/neg 26.42/neg 25.52/neg

Soybean meal (organic) 19.20/40,960 19.10/1920 20.10/800 22.55/60

Soybeans (conventional) 22.39/3840 24.18/40 28.86/neg 20.99/neg

Soybean meal(conventional) 19.75/7680 19.71/1920 19.76/5120 16.04/60

Lysine hydrochloride 18.54/1280 18.51/40 18.79/100 17.83/100

D-L methionine 22.17/2560 21.25/neg 25.03/neg 19.50/neg

Tryptophan 23.56/160 22.54/neg 25.30/neg 21.31/neg

Vitamin A 25.22/neg 22.66/neg 25.23/neg 26.85/neg

Vitamin D 22.28/3840 21.29/640 22.72/320 20.94/40

Vitamin E 28.54/neg 22.86/neg 26.15/neg 30.86/neg

Choline chloride 20.82/40 20.90/40 20.41/neg 20.77/neg

Rice hulls 24.73/neg 24.47/neg 25.93/neg 22.99/neg

Corn cobs 23.66/80 22.83/neg 24.68/neg 24.49/neg

Tetracycline 38/neg 38/neg 38/neg 38/neg

(−) control feed 38/neg 38/neg 38/neg 38/neg

Virus stock 17.15/245,760 18.86/30 21.19/neg 23.73/neg

Ingredient: Two 30 g replicates per ingredientMean Ct/FFU titer: Mean Ct value and FFU/mL across the 2 samples per ingredient

Dee et al. BMC Veterinary Research (2016) 12:51 Page 7 of 12

chloride. Viable PEDV was detected in piglets adminis-tered non-treated samples of choline chloride (Table 5).Affected animals displayed evidence of mild diarrhea, shedPEDV in feces and samples of small intestine were PCRand immunohistochemistry-positive at necropsy, withmicroscopic lesions of villous blunting, fusion and re-epithelialization [5]. All other samples were bioassay nega-tive. In regards to treated ingredients, LA-treated andMCFA-treated samples of soybean meal (conventionaland organic), lysine, vitamin D and choline chloride weretested. All piglets inoculated with the aforementioned LA-treated or MCFA-treated ingredients were determined to

Table 5 Summary of results of PCR (+)/VI (−) non-treated control fe

Ingredient Ct of inoculum C

Soybean-organic 25.52 n

Soybean-conventional 26.34 n

Vitamin A 26.86 n

Vitamin E 30.86 n

Rice hulls 22.94 n

Corn cobs 23.95 n

Tryptophan 21.31 n

D/L methionine 19.75 n

Choline chloride 20.79 p

(−) control >38 n

be non-infectious, as piglets remained clinically normalthroughout the testing period and all rectal swab and in-testinal samples were negative by PCR (Table 6).

Environmental dataFigure 2 provides a summary of the mean dailytemperature and % RH data recorded in the environ-mental chamber throughout the 37-day study period.For the purpose of statistical analysis, the 37-dayperiod was divided into 4 segments: Days 1–8, repre-senting time in China, days 9–25, representing timetravelling across the Pacific, days 26–32, representing

ed ingredients from batch 4 tested by swine bioassay

linical signs/rectal swabs PCR testing of small intestine

egative negative

egative negative

egative negative

egative negative

egative negative

egative negative

egative negative

egative negative

ositive positive

egative negative

Table 6 Summary of results of PCR (+)/VI (−) LA-treated or MCFA-treated control feed ingredients from batch 4 tested by swinebioassay

Ingredienta Treatment Ct of inoculum Clinical signs & rectal swabs PCR testing of small intestine

Soybean meal-conventional LA 31.44 negative negative

Soybean meal-organic LA 22.38 negative negative

Vitamin D LA 38 negative negative

Lysine LA 17.83 negative negative

Choline chloride LA 33.70 negative negative

Soybean meal-conventional MCFA 23.78 negative negative

Soybean meal-organic MCFA 17.75 negative negative

Vitamin D MCFA 20.60 negative negative

Lysine MCFA 20.33 negative negative

Choline chloride MCFA 21.24 negative negative

(−) control Saline 38 negative negativea: Ingredients were selected based on the recovery of viable PEDV in their non-treated equivalent samples

Dee et al. BMC Veterinary Research (2016) 12:51 Page 8 of 12

time spent in the San Francisco terminal and days33–37, time spent during transport from California toIowa with time in Des Moines (Table 7). In summary,mean temperature and mean % RH recorded duringthe San Francisco segment were significantly different(p = 0.0004 and p = 0.0025, respectively) from that re-corded during the other 3 segments.

Fig. 2 Mean daily temperature and % RH recorded during the 37-day studconditions recorded during the simulated period of December 23, 2012 to

DiscussionOver the course of less than 1 year (April 2013-February2014), several novel corona viruses have been detectedin the US including the PEDV prototype strain, thePEDV S INDEL strain and the porcine deltacoronavirus[14–16]. As of this writing, the route(s) of entry of thesepathogens has not been identified; however, phylogenetic

y period. This figure depicts the summary of the environmentalJanuary 28, 2013

Table 7 Descriptive summary of temperature and % RH data per the 4 segments of the 37-day study period

Temp C0 Segment 1 (China) Segment 2 (Pacific) Segment 3 (SF) Segment 4 (DSM)

1–8 DPC 9–25 DPC 26–32 DPC 33–37 DPC

Mean 5.5 6.0 8.4 3.9

95 % CI 4.3–6.7 5.2–6.8 7.1–9.6 2.5–5.4

SD 1.4 2.0 1.2 .13

Median 4.8 5.5 8.3 3.9

Range 4.2–7.8 3.9–9.4 6.3–10 3.9–4.2

% RH Segment 1 (China) Segment 2 (Pacific) Segment 3 (SF) Segment 4 (DSM)

1–8 DPC 9–25 DPC 26–32 DPC 33–37 DPC

Mean 71 63 84 76

95 % CI 63–79 58–69 75–92 66–87

SD 21.6 1.3 6.2 13.1

Median 70 63 83 70

Range 26–92 62–66 73–94 64–92

Dee et al. BMC Veterinary Research (2016) 12:51 Page 9 of 12

analyses support the possibility of a Chinese origin [11,14–16]. Since the initial detection of PEDV in the US,feed has been proposed as a possible route of entry andwidespread, rapid distribution throughout the country;however, objective data supporting this hypothesis werelacking. Therefore, the objectives of this study were todevelop a model to evaluate the transboundary risk ofPEDV-contaminated swine feed ingredients and test theeffect of two mitigation strategies during a simulatedtransport event from China to the US. This approachallowed us to design a study to successfully simulate theintroduction of PEDV into the US in feed ingredients;the first such objective data of its kind.The central component of the study was the develop-

ment of a model to accurately represent the conditionsof a trans-oceanic shipping event to determine whetherthe virus could actually survive the process. Key compo-nents of the model were the selection of the proper in-gredients, the calculation of an accurate timetable, andthe simulation of representative environmental condi-tions to challenge virus viability. In regards to ingredientselection, we utilized the US Government HarmonizedTariff Schedule to include products regularly importedto the US from China, 5 of which ended up harboringviable PEDV throughout the 37-day period: vitamin D,lysine hydrochloride, choline chloride, organic and con-ventional soybean meal. It was interesting to note thatviable virus was once again recovered from 3 ingredi-ents, (conventional soybean meal, lysine and cholinechloride). This was similar to our previous report [10],despite the fact that the 2 studies were conducted undervery different environmental conditions and that labora-tory personnel were blinded to sample identity/origin inboth studies. This level of repeatability strengthens theconclusions drawn from both studies, thereby increasing

the significance of these specific ingredients as potentialtransboundary risks.Besides the consistency of the results, these ingredi-

ents were interesting for several other reasons. Inregards to soybean meal, it was surprising to learn of thelarge quantity of soy products imported to the US fromChina, as the US is a major producer/exporter of soy-beans and soybean meal. Using Google to search forsources of both organic and conventional soy productsfrom China, we identified multiple soybean and soybeanmeal manufacturers in the eastern region of the countrysupplying product for agricultural use in bags, totes,containers or bulk quantities, including products desig-nated as “organic”, targeted specifically for organic live-stock feeding. Therefore, these data, in combinationwith our previous work demonstrating extended PEDVsurvival in soybean meal in cold climates [10] suggestthat soy-based ingredients could be considered a signifi-cant transboundary risk for pathogen transmission, in-cluding the organic variety. Another ingredient whichharbored viable virus throughout the 37-day transportperiod was lysine hydrochloride, a similar finding to thatreported in the 30-day wintertime study [10]. Lysine isan interesting ingredient as the volume imported to theUS from China is large (particularly in 2013) whichcould enhance the significance of this ingredient as a po-tential risk factor for transboundary spread of PEDV. Fi-nally, this study brought forth new information on therisk of PEDV survival in vitamins, specifically vitamin D.During the early stages of the PED epidemic in 2013, vi-tamins and trace minerals were proposed as a route ofviral entry, based on the volume of product importedfrom China. While viable PEDV was not recovered frompreviously tested vitamin/trace mineral mixes [10], itwas speculated that this may have been due to the

Dee et al. BMC Veterinary Research (2016) 12:51 Page 10 of 12

potentially caustic effect of the minerals on the virus.Therefore, we decided to investigate vitamins individu-ally, and while vitamin A and E did not support virus,results were different with vitamin D. Finally, as in theprevious study [10], choline chloride again proved cap-able of harboring viable PEDV over time, thereby raisingits significance as a possible transboundary risk factor.Along with ingredient selection, the second key com-

ponent of the model was the calculation of the shippingtimetable in order to truly represent this journey. UsingSearates.com, we developed an accurate estimate of timerequired to move from the various cities and portswithin the model. It did not, however, include time inport for preparation for shipping or clearance of cus-toms. Therefore, a 7-day period of time was added at thepoint of each port to better represent reality (T. Schmitt,APC Inc., personal communication, 2015), increasingthe calculated timetable from 23 to 37 days. This modifi-cation further challenged the viability of the virus,thereby strengthening the model.Finally, perhaps the most novel component of the

model was the development of the environmental curvewhich included actual temperature and % RH data fromocean containers travelling from Asia to the US [12].This interesting project was carried out by the technicalstaff of the Xerox Company and involved the use of dataloggers placed inside the actual containers, collectingreal time temperature and % RH data as they traveledaround the world. As one of the datasets involved ashipping event from Asia to the US in the months ofDecember and January, we used this information as thecenterpiece of our model and were able to reproducethe consistent trans-Pacific temperatures and % RH de-scribed in their publication along with the greater vari-ability observed on land. Combining this with historicaldata from the various cities and ports using WeatherUnderground, we were able to develop a representativetemperature and % RH curve, allowing us to programthe environmental chamber accordingly (Fig. 2). One in-teresting observation was the rapid reduction in stockvirus titer observed from the initial titer used to inocu-late batch 1 (mean titer = 245,760 FFU) to that detected8 days later in batch 2 (mean titer = 30 FFU) with theeventual loss of viability (batch 3) in the stock virus con-trol, despite storage in MEM, an ideal medium for virusstorage. This observation indicates that the environmen-tal conditions used in the study were rigorous, therebychallenging virus survival. This was most likely due tothe warmer temperatures and high % RH present in themodel, conditions quite unlike the winter time projectwhere long-term PEDV stock virus survival was ob-served [10]. We insured that ingredients would be ex-posed to the effect of the % RH by adding the holes inthe sides of each container model. It also suggests the

need for a supportive matrix to maintain viability in thecontainer during shipment and that under the correctconditions, PEDVis highly susceptible to environmentalpressures if left unprotected in its respective shippingcontainer. Finally, the fact that viable PEDV was recov-ered from only 5 of 14 ingredients suggests that individ-ual ingredient chemistry may provide a protective effect,potentially shielding virus from the environment andallowing it to survive and this may be particularly true ofthe meal form of soy versus the intact bean. In contrast,the high alkalinity of feed grade tetracycline (pH 13) ap-parently degraded the RNA immediately post-contamination, resulting in PCR-negative readings in all4 batches.In an effort to assess the potential for chemical mitiga-

tion of ingredient risk, two interventions, the liquid anti-microbial SalCURB® and the medium chain fatty acidblend were included in the design. While we had vali-dated the LA in previous studies [8, 10], this was ourfirst opportunity to assess the MCFA option. While theeffect of the LA on PEDV RNA degradation over thecourse of the study was significantly different than theresponse to MCFA, both products appeared to haveequivalent effect on virus viability. This outcome sup-ports the validity of chemical mitigation as a means toreduce the risk of PEDV in feed ingredients stored underconditions modeling the trans-oceanic voyage, as well asprovides options for treatment.In addition to an improved understanding of risks and

mitigation, other strengths of the study included an ex-perimental design where the only variables in the studywere the individual ingredients, and the presence or ab-sence of the treatment. Specifically, we inoculated equalamounts of each ingredient with an equivalent quantityof virus in an attempt to mimic published viral loads as-sociated with field cases of PEDV in feed. All ingredientswere stored in identical models of shipping containers,exposed to the same environment and tested in a singlelaboratory, involving consistent, trained personnel andvalidated assays for the detection of PEDV. We orga-nized samples in a manner to prevent cross-contamination by insuring that the individual samplestorage containers were never opened from the time ofPEDV inoculation until testing occurred at the labora-tory. While this may have resulted in some variability ofthe PCR assay since a new batch was submitted eachtime, the fact that our negative control samplesremained free of contamination throughout the entireproject validated these protocols. Finally, we used mul-tiple metrics (PCR, VI and bioassay) to document viralload across a total of 360 samples, exceeding our previ-ous work.However, as with all studies there were both strengths

and limitations. For example, we were limited to 14

Dee et al. BMC Veterinary Research (2016) 12:51 Page 11 of 12

ingredients primarily due to cost; however, the decisionto use similar ingredients across the 2 studies enhancedreplication and confidence in the data. While we werelimited in the number of replicates, each batch con-tained 2 replicates/ingredient/batch; resulting in 8 repli-cates per ingredient. Furthermore, these results werederived from gram quantities of feed and may not dir-ectly equate to the vast quantity of tonnage used in ac-tual swine production or feed manufacturing facilities.The bioassay protocol, while helpful at identifying ingre-dients containing low levels of virus was costly and diffi-cult to carry out. As we evaluated only one shippingroute over a defined period of time under a specific setof environmental conditions, this does not allow for ex-trapolation of results across other climates, other routesor over longer periods of time. Finally, as this was an ex-periment and did not take place in China, we could notprovide proof on how the contamination of ingredientscould actually take place. However, from our experiencesin the US, we could speculate that contamination of feedcould take place through the inclusion of contaminatedingredients or through contact with the virus in the en-vironment, either at the farm or at the mill [6, 10]. AsPEDV dispersion throughout the mill setting appears tobe widespread, contact with ingredients within a millingstorage facility seems logical [17].In closing, under the conditions of this study, we pro-

vided the first proof of concept data indicating that a sub-set of contaminated feed ingredients could successfullytransport live PEDV into the US using a novel trans-boundary shipping model. These are the first objectivedata suggesting that feed ingredients could potentiallyserve as vehicles for pathogen transfer between countries,as well as providing a plausible explanation regarding howPEDV was introduced to the US in 2013. This study alsoprovides insight on how the onset of organic farmingcould serve to elevate risk of pathogen introduction intothe US, secondary to the entry of organic soy products.This study also introduces a model which could enhancefurther transboundary research efforts, possibly employingsurrogate viruses (bovine viral diarrhea virus for swinefever virus or Seneca Valley virus for foot-and-mouth dis-ease virus) to test feed-related risks for other foreign ani-mal diseases, further the investigation of the risk oforganic ingredients, along with the continuing validationof mitigation strategies. With this information, it is hopedthat veterinarians, feed industry experts and governmentofficials will work together to develop a comprehensiveplan of ingredient management to reduce global risk. It istime for discussions on how to adequately treat feed, howto increase biosecurity at the port level, as well as whereto purchase biosecure ingredients, with the end result be-ing a lowered incidence of feed-related disease transmis-sion around the world.

ConclusionsThese results indicate the ability of PEDV to survive inspecific feed ingredients under modeled conditionssimulating shipment from China to the US. This is thefirst proof of concept suggesting that contaminated feedingredients could serve as transboundary risk factors forPEDV, along with the identification of effective mitiga-tion options.

Availability of supporting dataThe data set(s) supporting the results of this article is in-cluded within the article.

AbbreviationsPEDV: porcine epidemic diarrhea virus; PED: porcine epidemic diarrhea;Ct: Cycle threshold; PCR: polymerase chain reaction; ARW: Animal ResourceWing; MEM: minimal essential media; SBM: soybean meal; LA: liquidantimicrobial.

Competing interestsThe authors declare no competing or conflicting interests associated withthis study.

Authors’ contributionsSD: Developed study, co-wrote paper. CN: Developed study, provided swinenutrition expertise, reviewed paper. TC: Conducted molecular diagnostics,co-wrote paper. AS: Provided virological expertise at the laboratory level,co-wrote paper. JCH: Provided critical review and revising of paper. EN:Provided virological expertise at the laboratory level, co-wrote paper. CJ &RC: Provided MCFA product, resources, co-wrote paper. GP: Analyzed the USgovernment Harmonized Tariff Schedule website, reviewed paper. GS:Developed study, reviewed paper, mentored project. All authors read andapproved the final manuscript.

AcknowledgementsThe authors would like to recognize Dr. Michele Mucciante and the AnimalResource Wing team for their significant contributions to the success of thisstudy. The authors would also like to thank Kemin Industries, APC Inc andthe National Pork Board for providing the technical expertise, funding andin-kind resources necessary to complete this project.

Author details1Pipestone Applied Research, Pipestone Veterinary Services, 1300 Box188Hwy 75 S, Pipestone, MN 56164, USA. 2Animal Disease Research andDiagnostic Laboratory, South Dakota State University, Brookings, SD, USA.3Department of Grain Science, Kansas State University, Manhattan, KS, USA.4Center for Animal Health and Food Safety, University of Minnesota, St. Paul,MN, USA.

Received: 26 October 2015 Accepted: 4 March 2016

References1. Saif LJ, Pensaert MB, Sestak K, Yeo S, Jung K. Coronaviruses. In: Zimmerman

JJ, Karriker LA, Ramierez A, Schwartz KJ, Stevenson GW, editors. Diseases ofSwine. 10th ed. Ames: Wiley and Sons; 2012. p. 501–24.

2. Chen Q, Ganwu L, Stasko J, Thomas JT, Stensland WR, Pillatzki AE, GaugerPC, Schwartz KJ, Madson D, Yoon KJ, Stevenson GW, Burrough ER, HarmonKM, Main RG, Zhang J. Isolation and characterization of Porcine EpidemicDiarrhea Viruses associated with the 2013 disease outbreak among swine inthe United States. J Clin Microbiol. 2014;52:234–43.

3. American Association of Swine Practitioners, National Pork ProducersCouncil and USDA Center for Epidemiology and Animal Health. PEDVInvestigation 2013. http://www.aasv.org. Accessed 22 July 2015.

4. Goyal S. Interventions to control PEDV in feed and feed ingredients. http://www.pork.org/pedv-2014-research/pedv-feed. (2014). Accessed 23 March2015.

Dee et al. BMC Veterinary Research (2016) 12:51 Page 12 of 12

5. Dee S, Clement T, Schelkopf A, Nerem J, Knudsen D, Hennings J, Nelson E.An evaluation of contaminated complete feed as a vehicle for porcineepidemic diarrhea virus infection of naïve pigs following consumption vianatural feeding behavior: Proof of concept. BMC Vet Res. 2014;10:176.

6. Pascik J, Berhane Y, Ojkic D, Maxie G, Embury-Hyatt C, Swekla K, Handel K,Fairles J, Alexandersen S. Investigation into the role of potentiallycontaminated feed as a source of the first-detected outbreaks of porcineepidemic diarrhea in Canada. Transbound Emerg Dis. 2014;61:397–410.

7. Schumacher L, Woodworth JC, Zhang J, Gauger PC, Chen Q, Welch M,Salzebrenner H, Thomas J, Main R, Dritz S, Cochrane R, Jones C.Determining the minimum infectious dose of PEDV in a feed matrix.Proceedings of the 2015 Midwest Animal Science Conference Des Moines,IA, USA, Abstract. 2015;160.

8. Dee S, Neill C, Clement T, Hennings J, Nelson E. An evaluation of a liquidantimicrobial (SalCURB®) for reducing the risk of porcine epidemic diarrheavirus infection of naïve pigs during consumption of contaminated feed.BMC Vet Res. 2014;10:220.

9. Cochrane R. Evaluation of chemical mitigation of PEDV in feed andingredients. ADSA-ASAS Midwest Meeting, Abstract. 2015;11670.

10. Dee SA, Neill C, Clement T, Singrey A, Christopher-Hennings J, Nelson E. Anevaluation of PEDV survival in individual feed ingredients in the presence orabsence of a liquid antimicrobial. Porcine Health Management. 2015;1:9.

11. Huang YW, Dickerman A, Pineyro P, Long L, Fang L, Kiehne R, Opressnig T,Meng XJ. Origin, evolution and genotyping of emergent porcine epidemicdiarrhea virus strains in the United States. mBio. 2013;4:00737–13.

12. Leinberger D. Temperature and humidity in ocean containers. Technicalreport: Xerox Corporation; 2006. http://www.ista.org/pages/resources/TempRH.php.

13. Okda F, Liu X, Singrey A, Clement T, Christopher-Hennings J, Nelson EA,Lawson S. Development of an indirect ELISA, blocking ELISA, fluorescentmicrosphere immunoassay and fluorescent focus neutralization assay forserological evaluation of exposure to North American strains of porcineepidemic diarrhea virus. BMC Vet Res. 2015;11:180.

14. Wang L, Byrum B, Zhang Y. Detection and genetic characterization ofdeltacoronavirus in pigs Ohio US. Emerg Infect Dis. 2014;20:1227–30.

15. Vlasova A, Marthaler D, Wang Q, Culhane M, Rossow K, Rovia A, Collins Jand Saif L. Distinct characteristics and complex evolution of porcineepidemic diarrhea virus strains in North America: May 2013-February 2014.Emerg Infect Dis. 2014;20:1620–8.

16. Wang L, Byrum B, Zhang Y. New variant of porcine epidemic diarrhea virus,United States. Emerg Infect Dis. 2014;20:917–9.

17. Schumacher LL, Cochrane RA, Evans CE, Kalivada JR, Woodworth JC, StarkCE, Jones CK, Main RG, Zhang J, Dritz SS, Gauger PC. Evaluating the effect ofmanufacturing porcine epidemic virus diarrhea virus-contaminated feedand subsequent feed mill environmental surface contamination. KansasAgricultural Experiment Station Research Report. 2015, Vol 1, Issue 7.

• We accept pre-submission inquiries

• Our selector tool helps you to find the most relevant journal

• We provide round the clock customer support

• Convenient online submission

• Thorough peer review

• Inclusion in PubMed and all major indexing services

• Maximum visibility for your research

Submit your manuscript atwww.biomedcentral.com/submit

Submit your next manuscript to BioMed Central and we will help you at every step: